A probabilistic process learning approach for service composition in cloud networks

We present a formal probabilistic framework for process learning to compose service specific overlays (SSO) in cloud networks. The approach provides a learning mechanism that relies on previous composition results to build service composition process models that can be adopted for future composition requests. The process is then translated into a workflow-net to provide guaranteed delivery of requested cloud media services to clients. A mathematical merge technique is also presented to converge multiple process threads into a single composed process. We provide simulation results to show that our approach can adequately establish sound composition paths in a timely manner.